摘要
针对空间遥感任务中,光电探测器配载的斯特林制冷机制冷时长的预测问题,提出了一种使用自回归积分滑动平均模型(ARIMA)与人工神经网络(ANN)模型相结合的预测方法。该方法采用时间序列的经典模型,将时间序列分解为趋势项与波动项的加和,采用ARIMA和ANN模型分别进行趋势项预测和波动项预测,并根据斯特林制冷机的间歇性工作模式特点对ARIMA模型进行改进。最后,通过在真实数据上与其他方法的对比实验,验证了本文提出方法的有效性。
A new method combining autoregressive integratedmoving average （ARIMA） model and artificial neuralnetwork （ANN） model is presented for the duration prediction of the Stifling cryocooler applied in space remotesensing missions. The stirling cryocooler is equipped with electrophotonic detector. A classic model based on timeseries is used to decompose the time series into trend item and oscillation item. For prediction, ARIMA model isused to predict the trend item and ANN model is used to predict the oscillation item, while ARIMA model is im-proved based on the intermittent workingmode of the Stifling cryocooler. Finally,the validity of this method is veil-fled by applying to the real data and comparing the resuhs with other methods.
出处
《应用科技》
CAS
2015年第5期61-66,共6页
Applied Science and Technology
基金
载人航天空间应用任务地面支持项目(Y2140411SN)
